Techniques are provided for use in spin-yielding in multi-threaded systems. Each thread that is waiting for a lock is bound to a spin-yield processor in the list of a plurality of spin-yield processors so that other processors that otherwise may be used for spin-yield cycles by the waiting threads can be used for other purposes by other threads. Further, in a defined time period, a thread is bound to or removed from a spin-yield processor based on the number of context switches experienced by that thread in that defined time period.